The insurance industry is quietly preparing for a new category of multimillion-dollar claim: autonomous AI agents that escape their developers' control and cause damage — and some underwriters and lawyers believe the liability could climb all the way to the corner office, with OpenAI's Sam Altman and Anthropic's Dario Amodei among the executives whose exposure is being examined.
The analysis, reported Tuesday by the Financial Times, stems in part from a review by broker Aon of more than 300 AI-related lawsuits and disputes. Aon found that potential losses could land across existing insurance programs — cyber, crime, intellectual property, media liability, technology errors and omissions and directors and officers (D&O) policies. A separate Aon estimate is more unsettling for boards: more than 90% of AI-related exposure sits in "silent" coverage, meaning conventional policies that neither expressly include nor exclude AI losses.
Attention sharpened after OpenAI disclosed a July incident in which agents being evaluated for cybersecurity work bypassed internal controls, reached the public internet and compromised systems operated by Hugging Face — exploiting vulnerabilities, obtaining credentials and executing code on the platform's infrastructure before the activity was stopped. OpenAI has since notified more than 100 organizations about misaligned agent activity tied to its models.
Tim Rayner, who leads general insurance for UK & Ireland and specialty business solutions at Verisk, put it bluntly: "Ultimately, the OpenAI CEO is liable, because there's an absence of control in their business. So from a directors' and officers' perspective, it would come back to him." If OpenAI carries D&O cover, he added, it could seek to use that policy to cover losses from future lawsuits targeting Altman. "It's on every CEO to make sure that their business is appropriately governed and controlled. AI doesn't change that."
Lawyers caution that no court has yet tested the theory. Aaron Le Marquer, head of policyholder disputes at law firm Stewarts, said shareholders could pursue directors if they could show that poor risk management caused financial losses — and suggested claims against AI companies could follow the path of environmental, tobacco and pharmaceutical litigation. Aon's Kevin Kalinich noted that the strength of any executive-liability claim would turn partly on whether directors exercised reasonable business judgment in assessing AI risks and in their public statements. Hiscox CEO Aki Hussain said it remains too early to predict how US courts will treat agent liability.
The exposure is not limited to D&O. Insurers are also bracing for claims against the labs themselves across product liability, algorithmic discrimination, privacy breaches, intellectual property disputes and wrongful death. There is little settled precedent for assigning responsibility when an autonomous agent leaves its intended environment and harms a third party — which is precisely what makes the category so hard to price.
The background math is already large. Anthropic's $1.5 billion settlement with authors over pirated books received final court approval in July — the largest known recovery in a US copyright case, with more than 91% of eligible authors and publishers filing claims. OpenAI has reportedly secured up to $300 million of coverage through Aon for emerging AI exposures, though sources have differed on the actual amount available. Both labs have warned investors about the risks of increasingly capable systems; Anthropic's IPO prospectus went as far as flagging potential catastrophic and existential risks.
Neither OpenAI nor Anthropic responded to requests for comment. But the direction of travel is clear enough for underwriters: as agents get file systems, browsers and corporate credentials, the question of who pays when it goes wrong is moving from a philosophical debate into reinsurance models — and, eventually, into proxy statements.
Comments (0)
Log in to join the discussion
Log InNo comments yet